| Message ID | 20180328141205.2954245-1-arnd@arndb.de |
|---|---|
| State | New |
| Headers | show |
| Series | time: export nsec_to_clock_t | expand |
On 三, 2018-03-28 at 16:11 +0200, Arnd Bergmann wrote: > nsec_to_clock_t was traditionally used only in the core kernel, now > we > have a sysfs file that needs it from a loadable module, causing a > link-time error: > > ERROR: "nsec_to_clock_t" [drivers/thermal/thermal_sys.ko] undefined! > > This exports the function the same way that we do for related > interfaces. > > Fixes: 96cea33badc5 ("thermal: Add cooling device's statistics in > sysfs") > Signed-off-by: Arnd Bergmann <arnd@arndb.de> Thanks for the fix. can I take this patch through thermal tree? thanks, rui > --- > kernel/time/time.c | 1 + > 1 file changed, 1 insertion(+) > > diff --git a/kernel/time/time.c b/kernel/time/time.c > index 6fa99213fc72..97a262531f68 100644 > --- a/kernel/time/time.c > +++ b/kernel/time/time.c > @@ -768,6 +768,7 @@ u64 nsec_to_clock_t(u64 x) > return div_u64(x * 9, (9ull * NSEC_PER_SEC + (USER_HZ / 2)) > / USER_HZ); > #endif > } > +EXPORT_SYMBOL(nsec_to_clock_t); > > u64 jiffies64_to_nsecs(u64 j) > {
On Thu, 29 Mar 2018, Zhang Rui wrote: > On 三, 2018-03-28 at 16:11 +0200, Arnd Bergmann wrote: > > nsec_to_clock_t was traditionally used only in the core kernel, now > > we > > have a sysfs file that needs it from a loadable module, causing a > > link-time error: > > > > ERROR: "nsec_to_clock_t" [drivers/thermal/thermal_sys.ko] undefined! > > > > This exports the function the same way that we do for related > > interfaces. > > > > Fixes: 96cea33badc5 ("thermal: Add cooling device's statistics in > > sysfs") > > Signed-off-by: Arnd Bergmann <arnd@arndb.de> > > Thanks for the fix. > can I take this patch through thermal tree? Well, the question is what's the point of that? Why does a new interface expose time in state in clock_t instead in normal time units, e.g. nsec/usec/msec ? That makes no sense whatsoever. Thanks, tglx
On 29-03-18, 21:11, Thomas Gleixner wrote: > On Thu, 29 Mar 2018, Zhang Rui wrote: > > On 三, 2018-03-28 at 16:11 +0200, Arnd Bergmann wrote: > > > nsec_to_clock_t was traditionally used only in the core kernel, now > > > we > > > have a sysfs file that needs it from a loadable module, causing a > > > link-time error: > > > > > > ERROR: "nsec_to_clock_t" [drivers/thermal/thermal_sys.ko] undefined! > > > > > > This exports the function the same way that we do for related > > > interfaces. > > > > > > Fixes: 96cea33badc5 ("thermal: Add cooling device's statistics in > > > sysfs") > > > Signed-off-by: Arnd Bergmann <arnd@arndb.de> > > > > Thanks for the fix. > > can I take this patch through thermal tree? > > Well, the question is what's the point of that? > > Why does a new interface expose time in state in clock_t instead in normal > time units, e.g. nsec/usec/msec ? The initial patch was displaying the values in msec, but eduardo asked the units to be consistent with how cpufreq shows it, i.e. clock_t. https://lkml.kernel.org/r/20180112174606.GA11076@localhost.localdomain Anyway I will resend the patch and move back to msec. Thanks. -- viresh
diff --git a/kernel/time/time.c b/kernel/time/time.c index 6fa99213fc72..97a262531f68 100644 --- a/kernel/time/time.c +++ b/kernel/time/time.c @@ -768,6 +768,7 @@ u64 nsec_to_clock_t(u64 x) return div_u64(x * 9, (9ull * NSEC_PER_SEC + (USER_HZ / 2)) / USER_HZ); #endif } +EXPORT_SYMBOL(nsec_to_clock_t); u64 jiffies64_to_nsecs(u64 j) {
nsec_to_clock_t was traditionally used only in the core kernel, now we have a sysfs file that needs it from a loadable module, causing a link-time error: ERROR: "nsec_to_clock_t" [drivers/thermal/thermal_sys.ko] undefined! This exports the function the same way that we do for related interfaces. Fixes: 96cea33badc5 ("thermal: Add cooling device's statistics in sysfs") Signed-off-by: Arnd Bergmann <arnd@arndb.de> --- kernel/time/time.c | 1 + 1 file changed, 1 insertion(+) -- 2.9.0